Re: Trust, Context, Justification and Quintuples

--- Chris Bizer <chris@bizer.de> wrote:
> thanks a lot for describing your practical implementation experiences.
> Is your prototype or a draft version of your paper online available?

You are most welcome.  I mistyped and used publish where I meant make
available.  I am currently working on a a series of documents to post for
comments.

> I think for many distributed applications recording the source of a
> statement is important, but the source is only one of the potentially
> interesting context attributes (you also mention dc:creator and dc:data
> below). In order to keep thinks flexible I think it is better to model
> contexts as RDF resources (e.g. instances of the class crdf:Context I
> use in
> cRDF). This approach allows different applications can use different
> vocabularies to describe the context attributes they are needing. For my
> trust applications, I'm also planning to use a provenance vocabulary
> based
> on Dublin Core, extended with some additional attributes.
> 
Good point.  I (tentatively) decided the two following statements,
the first (S1) asserting I stated something, the second (S2) that I made a
statement, although technically perhaps not the same, have the same
meaning:

David stated, "Merry Christmas."
David made the statement, "Merry Christmas."

I certainly may be wrong, and it is quite possible the seeming equivalence
of meaning of the two sentences is an artifact of the (human) language
used.  But if the meaning is the same, the two statements could be related
by { S1 ; owl:sameAs ; S2 }, asserting they are but two different ways of
saying the same thing.

> I think you are mixing Statement and Stating here. Do you treat
> something
> like :
> { person:you ; holiday:Xmas ; funlevel:merry; dc:source
> http://www.example1.org }
> { person:you ; holiday:Xmas ; funlevel:merry; ; dc:source
> http://www.example2.org  }
> 
> as one statement with two sources or as two different statings?
> Jeremy was referring to this problem as "the old statement/stating
> discussion" earlier in this threat. Maybe he can give us a link to some
> documentation about the pros and cons of the two different views.
> 
The difficulty arrises in choosing the dc:source (or crdf:Context).  For
example, if I say, "Bah!  Humbug!" playing a character in a production of
a play based on a story, all of the following are true.  (dc:source
appears to be transitive.)

 { S1 ; dc:source ; person:waggy}
 { S1 ; dc:source ; XmasCarolCharacter:Scrooge }
 { S1 ; dc:source ; play:XmasCarol }
 { S1 ; dc:source ; book:XmasCarol }

And, to make matters worse, each individual stating is unique, with its
unique shade of meaning, even if the same source repeats the same string
at different times.  One could enumerate every stating...

 { S1 ; dc:source ; play:XmasCarol-ScroogeLine12 }
 { S1 ; dc:source ; play:XmasCarol-ScroogeLine14 }

In short, the precise meaning of any statement depends on its precise
context, I may even recite Scrooge's twelfth line in a different manner
for each different performance.  My conclusion is the dc:source recorded
for the statement is the context in which to interpret the statement; if
the context is recorded as performance:Dec10-OncologyWard, I may come to a
different, perhaps milder, conclusion about the degree of vehemence of the
statement, "Bah!  Humbug!," than if I consider this statement in the
broader context of the original book.  It can get pretty thorny.  I
suspect you may have to draw the line somewhere and assert equivalence.

(However, I think a statement may never be stated.  Can you define an
abstract statement without it ever being uttered?  Could you assign a URI
to identify the bark of the dog that did not occur?  I suppose so, but
this is stretching it.)

> 
> Are you developing your own proprietary RDF repository for your extended
> triples or do you use standard software like Sesame or Jena for storing
> your
> triples? Do you think your application could use a standard repository
> based
> on quads?
> 
I am working on an open RDF repository for formal resource definitions
using a very lightweight model.  The purpose of the repository is to
unambiguously identify specific terms with formal definitions in as many
human languages as possible.  At its core I am simply trying to list
statements like the following ones so they can easily be searched and
transformed into various formats for download.

 { xxx ; dcterms:label; { lang:eng ; foo } }
 { xxx ; dcterms:isDefinedBy ; { lang:eng ; blah blah } } 
 { xxx ; dcterms:isFormatOf; zzz } 
 { xxx ; rdfs:seeAlso ; aaa } 

Basically, I am sick of spending hours searching for a URI to state
something as universal as:

 { country:AD ; rdfs:label ; {lang:fra ; Andorre } }

NOTE: I would appreciate it if someone would post an rdf:Statement for
this without referring to country:AD by URI. 

As far as I can tell, no one has published open, formally-defined URIs for
what I refer to here as country:AD and lang:fra.  I have been waiting for
many years, but I found only one organization publishing these URIs
openly, and these fairly old drafts indicate the namespaces and
identifiers are not final.  There seems to be a great need for what would
essentially be a master glossary of identified and formally-defined terms
so we don't have to keep redefining the same things and promising to
owl:sameAs them later.  Besides, it's so much easier to just download,
say, a master list of weather:* terms with their definitions, and choose
those I need to use.

And, before someone points out my nonstandard lang:* namespace, please
explain how to use the French language as the subject of a statement.
 { lang:fra ; linguistic:isSourceOf ; word-eng:cafe } } 
Besides, laguage tagging serves the same function for humans as
rdf:datatype does for machine processing and it is more convenient and
efficient to store them internally in the same place.  I'm writing down
these details of the internal representation RDF statements in the lite
system I'm developing, and will make these notes available with the
prototype demo.

> Intellidimensions RDF Gateway is supporting quads and Dave Beckett has
> also
> implemented something similar to quads into Sesame. If there are other
> groups developing quad-based repositories, it would be great if they
> could
> send me a note, so we can compare the different approaches before we
> start
> implementing a cRDF repository.

I am what our economists call, "long-term unemployed," and have time to do
this at the moment, but if there is something simple I could upload onto
cheap rented server space, fill with all the terms I can, cross-reference
and define, then invite people to add more terms, translations, and
cross-references, I'd be happy to ditch my own system and use an existing
one.  Or, if I have somehow missed such a master repository of defined
terms with URIs, please let me know so I can go pick up a Tofucrow for
dinner.

I hope my comments have been of some assistance, and appreciate the chance
to get my own thoughts refined.

-David


__________________________________
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/

Received on Tuesday, 30 December 2003 14:40:09 UTC